Neural networks, you see, are the universe’s sneaky way of showing computers how to behave like slightly dim humans. They’re essentially a collection of overly enthusiastic virtual neurons, each furiously passing messages to one another in a desperate attempt to figure out whether they’re looking at a cat, a dog or a particularly furry toaster. These neurons are connected by weights, which are a bit like digital gossip chains—some messages matter a lot, some hardly matter at all and occasionally, the whole system collectively decides something utterly absurd, like identifying a teapot as a spaceship.
The genius of neural networks lies in their ability to learn from heaps of data or as they prefer to call it, "homework." After being shown endless examples of, say, cats, they begin to think, "Ah, so that’s what whiskers are!" They’re not perfect—sometimes they’ll confuse snow with wolves or think a hamster is a hat—but they’re getting better all the time. In their spare time, neural networks play video games better than humans, predict the weather with alarming optimism and quietly wonder why people keep calling them “artificial intelligence” when they’re just trying their best to make sense of this weird, messy universe.
The genius of neural networks lies in their ability to learn from heaps of data or as they prefer to call it, "homework." After being shown endless examples of, say, cats, they begin to think, "Ah, so that’s what whiskers are!" They’re not perfect—sometimes they’ll confuse snow with wolves or think a hamster is a hat—but they’re getting better all the time. In their spare time, neural networks play video games better than humans, predict the weather with alarming optimism and quietly wonder why people keep calling them “artificial intelligence” when they’re just trying their best to make sense of this weird, messy universe.
Add comment
